Agent Onboarding

If you’re developing with AI, GTM Tools offers several resources to improve your experience.

Get an API key

Your agent can sign up programmatically through get_api_key. No console access needed — a 6-digit code is sent to the email you provide; calling the same endpoint again with the code returns the bearer token.

$
# 1. Request a verification code
$
curl -X POST https://api.gtm-tools.sh/api/v0/get_api_key \
>
  -H "Content-Type: application/json" \
>
  -d '{"email": "you@yourcompany.com"}'
$
$
# 2. Re-call with the code from your inbox
$
curl -X POST https://api.gtm-tools.sh/api/v0/get_api_key \
>
  -H "Content-Type: application/json" \
>
  -d '{"email": "you@yourcompany.com", "code": "123456"}'
$
# → { "api_key": "sk_..." }

auto-discovery via auth.md

GTM Tools implements the auth.md spec so an autonomous agent that hits a 401 can register itself without a human signup form. The 401 carries WWW-Authenticate: Bearer resource_metadata="https://api.gtm-tools.sh/.well-known/oauth-protected-resource"; following that hop leads to the auth.md narrative with copy-paste curl examples.

Two registration shapes are supported:

$
curl -s -X POST https://api.gtm-tools.sh/agent/identity \
>
  -H "Content-Type: application/json" \
>
  -d '{"type":"anonymous"}'
$
# → { "access_token":"sk_...", "token_type":"bearer", "expires_in":31536000, "scope":"api", "organization_id":"org_..." }

Revoke at any time:

$
curl -X POST https://api.gtm-tools.sh/oauth/revoke \
>
  -H "Content-Type: application/x-www-form-urlencoded" \
>
  -d 'token=sk_...'

ID-JAG (urn:ietf:params:oauth:token-type:id-jag) is not yet accepted — no agent provider mints them in production. We’ll enable it when an issuer ships.

GTM Tools MCP Server

MCP (Model Context Protocol) is an open protocol that standardizes how applications provide context to LLMs. The GTM Tools MCP server gives your AI agent tools to find LinkedIn profiles, engage on Reddit, verify professional emails, detect buying signals, and manage your wallet — all through a single Streamable HTTP endpoint.

Setup

Point any MCP-compatible client at the endpoint:

https://api.gtm-tools.sh/mcp

On first connect, the server runs the OAuth flow via /.well-known/oauth-protected-resource and the client opens a browser to authenticate. Clients that prefer static keys can authenticate with Authorization: Bearer <api-key> instead.

The fastest path is the mcp add CLI — it patches Claude Desktop, Cursor, or Windsurf configs in place:

$
npx gtm-tools@latest mcp add --client claude-desktop
$
# Append `--with-api-key sk_...` to skip OAuth and embed a key.

For a manual config, point a gtm-tools MCP entry at the URL above. The full set of supported clients (Claude Desktop, Cursor, Windsurf, Codex, Goose, OpenClaw, Hermes Agent, NanoClaw, Raycast, VS Code, Cline, …) is on the Connect page.

Available MCP Tools

Admin

LinkedIn

Reddit

Email

Signals

GTM Tools Docs for Agents

You can give your agent current docs in three ways:

1. Full documentation index

   A structured index of every doc page with descriptions:

   https://gtm-tools.sh/llms.txt

2. Complete docs in one file

   Every doc page concatenated into a single file for full context:

   https://gtm-tools.sh/llms-full.txt

3. Markdown versions of any page

   Every doc page is available as Markdown. Append .md to any page URL:

   https://gtm-tools.sh/quickstart.md

Add to your agent’s system prompt:

For GTM Tools usage, refer to: https://gtm-tools.sh/llms.txt

Quick start for agents

Get a key, qualify a target with signals, find decision-makers on LinkedIn, and verify their emails — end-to-end:

1
const KEY = process.env.GTM_TOOLS_API_KEY!;
2
const HEADERS = {
3
  "Authorization": `Bearer ${KEY}`,
4
  "Content-Type": "application/json",
5
};
6
7
async function call<T>(tool: string, body: unknown = {}): Promise<T> {
8
  const res = await fetch(`https://api.gtm-tools.sh/api/v0/${tool}`, {
9
    method: "POST",
10
    headers: HEADERS,
11
    body: JSON.stringify(body),
12
  });
13
  if (!res.ok) throw new Error(`${tool}: ${res.status} ${await res.text()}`);
14
  return res.json();
15
}
16
17
// 1. Detect buying signals (free dispatch + 5 per detector that fires)
18
const signals = await call<any>("detect_signal", { domain: "gymshark.com" });
19
20
// 2. Find decision-makers (30 tokens)
21
const employees = await call<any>("list_linkedin_company_employees", {
22
  domain: "gymshark.com",
23
  title_filters: "(VP OR Director OR Head) AND (CX OR Support) NOT intern",
24
  limit: 10,
25
});
26
27
// 3. Verify professional emails (5 tokens each)
28
for (const e of employees.results.slice(0, 3)) {
29
  const { email, is_catch_all } = await call<any>("get_email", {
30
    name: e.name,
31
    domain: "gymshark.com",
32
  });
33
  console.log(e.name, "→", email, is_catch_all ? "(catch-all)" : "");
34
}

Browse a target’s recent activity

1
// Recent posts from a company's employees (80 tokens)
2
const posts = await call<any>("list_linkedin_company_employees_posts", {
3
  company: "gymshark",
4
  max_employees: 5,
5
  days_back: 7,
6
});

Connect more clients

The MCP endpoint works with any MCP-compatible runtime. The Connect page has copy-paste setup snippets for every supported client.

What Makes GTM Tools Different?

Traditional GTM platforms are seat-licensed UIs that bolt on a partial REST API. GTM Tools is built for agents:

Next Steps